Channel Estimation For Multi-Antenna Mobile Communications Based On Novel Training Sequences

Compared with single-input-single-output (SISO) communication system, multiple-input-multiple-output (MIMO) communication system is a wireless transmission scheme which uses multi-antenna array in transmitter and receiver. It has been proved that, compared with SISO system, MIMO system has much higher capacity which is approximately linearly increased with the number of transmitter antennas when the number of receiver antnennas is larger than that of transmitter antennas, thus breaks the traditional Shannon capacity limit. Nowadays, MIMO technique has become a key technology for the next generation mobile communication. For MIMO mobile communications, channel estimate is requisite. The way to estimate channel and the property of training sequences play a very important role in reducing the system complexity and improving the channel estimate performance. In this thesis, the channel estimate for MIMO system and related techniques based on novel training sequences are investigated.First of all, the channel estimate performance using implicit training sequences will be deteriorated by direct-current (DC) offset and data interference. In this thesis, a new MIMO channel estimate scheme based on zero correlation zone (ZCZ) implicit training sequences with balance property is proposed. Compared with traditional scheme using implicit optimum channel independent (OCI) training sequences, the new scheme has the advantages of DC offset-free performance without introducing extra complexity, lower peak-average-power ratio (PAPR), lower system complexity induced by binary sequence, and the new scheme is suitable for both SISO and MIMO systems.Next, an improved scheme for MIMO channel estimate using implicit ZCZ training sequences is proposed. The channel estimate performance will not be affected by DC-offset and data interference because the transmit data is the sum of training sequence and data symbol, where the circular mean is removed beforehand. In this process, it should be noted that the DC-offset is not estimated. However, the ultimate goal of communications is to detect and recover the transmitted data, therefore, the DC-offset should be removed before equalization.. Thus, the combined channel estimation and DC-offset estimation is studied, and it is shown that ZCZ sequences with balance property are optimal in both cases, and a lower performance bound based on least squared (LS) is derived.In the aforementined scheme, enough cyclic-prefix (CP) is added to the data block in order to avoid inter-block-interference (IBI) and to simplify channel estimation and equalization. The disadvantage is that the CP is redundant (non information-bearing overhead). In this thesis, a new formula for SISO/MIMO channel mean square error (MSE) estimate is derived in the absence of CP, and it is shown that the resultant MSE is not significantly altered by lacking of CP in the case of larger number of training sequence period and lower signal-noise-ratio (SNR).Then, based on time-domain analysis, a closed-form estimation variance as a function of sequence length, family size and zero correlation zone (ZCZ) is derived for MIMO systems employing ZCZ (explicit) training sequences. The performance of the least squares (LS) and a scaled LS (SLS) approaches to the channel estimation are investigated. Analysis and simulation results show that the ZCZ training sequence set outperforms other training sequence sets (m sequences, pseudo-random sequences), and the SLS method performs better than LS method at the price of more known prior information.It is proved analytically that the repeated phase-rotated Chu (RPC) sequences which are the optimal training sequences for MIMO system based on frequency-domain analysis are a special case of ZCZ sequences, its cross-correlation value is zero at all shifts, and its auto-correlation value is non-zero only at several shifts.Subsequently, a new scheme which combines space-time block-coding (STBC) based on Alamouti-like scheme and the LS channel estimation using optimal training sequences in MIMO-SCFD systems with CP (MIMO-STBC-SCFD) is proposed. With two transmit antennas, based on lower bound for LS channel estimation, it is shown that the periodic complementary set (PCS) is optimal over frequency-selective fading channels. Compared with the normal scheme without STBC, 3dB MSE performance gain and fewer restrictions on the length of channel impulse response are demonstrated.The requirement on the optimal training sequences for MIMO-STBC system channel estimation is derived based on time-domain analysis. It is shown that the optimality requirement relates to the undermentioned Z-periodic complementarity, and the desired optimal taining sequences, are in fact a class of Z-periodic complementary sequences which are an extension of L-perfect sequences.Finally, a new concept, called Z-periodic/aperiodic complementary binary sequences whose complementary crosscorrelation values and autocorrelation sidelobes are all zeros within a certain zone, called zero correlation zone, is proposed. Obviously, the new Z-complementary sequence sets include the conventional periodic/aperiodic complementary sequence sets as special cases. In this thesis, constructions and conjectures are also presented, including Z-complementary sequence sets and their mates.